Your statement still has some issues that need unpacking. “Barring accidents, the most fit individual will be selected” is not a statement I can assign a meaning to in this framework. I would say that, including accidents, the most fit individual is more likely to reproduce. Success at reproduction is a random process, i.e. it is nothing but accidents; all fitness tells you is how likely an individual is to have more of the successful accidents. There’s no deterministic process that goes on in the absence of accidents.
As is happens, we’re typically more interested in the fate of fitter traits, rather than fitter individuals. We can say, for example, that a newly arisen trait (specifically an additive trait, i.e. neither dominant nor recessive) that gives a 1% fitness advantage has a 2% probability of spreading throughout the population. That’s an unusually large selective advantage, and you see that it’s still easily swamped by the noise of random reproductive success. That’s why viewing success as the normal result – the thing that happens barring accidents – is not likely to give you a good picture of the situation. Fitness is usually a modest bias in a stochastic process.
Even with this clarification, though, there is an additional, subtler complication. If we agree to define the fitter trait as the one that causes the individual to have more offspring on average, then natural selection is the idea that fitter traits will be more likely to increase in frequency than less fit traits (note that this is again a propensity), all other things being equal. The kicker is that sometimes all other things aren’t equal. It sometimes happens that the less fit variant also has the effect of increasing its own inclusion in gametes, so that it’s more likely to be passed on. (Cartoon example: sperm that have the variant kill sperm that don’t have it.) In this case, the less fit trait can still spread and take over the population. To properly handle this kind of situation, you have to treat natural selection as operating on multiple levels, i.e. both the genetic and the organismal level. So a single simple statement about natural selection isn’t adequate.
It’s useful in a variety of contexts. Often it’s estimated retrospectively. It is of some biological interest, for example, that lactose tolerance provided a very large fitness advantage. It’s surprising and puzzling that this simple trait should be such an outlier in selection strength, and scientists like surprises and puzzles.
For another kind of use, consider this paper, which uses a rough guess at the fitness cost of a disease to give practical guidance on how to design genetic association studies. Such association studies are a big deal in the world of human disease genetics at present.